$b^3-a^2b+2a \geq 0$
Given: $(a,b) \in \mathbb{N}$ and a>b. Find when this holds. This is an inequality I stumbled upon while I was surfing, and like any other day, I sat down trying to think about it. In general, playing with it gives me that $b \geq 3$ "guarantees" no value of $a$. Although proving the statement formally seems to take me forever. Initially, I tried to bring it to quadratic form in $a$, and if $b$ behaves like a constant, then we can see $P(a) \geq 0$ which imposes $D \leq 0$, but that implies $b^4+1 \leq 0$ which is absurd in the field we are working on. This somehow convinced me to try working on the cubic which doesn't seem to be factorizable either.
I would love a solution that could include the remedy and why my approach didn't work.
Thank you.
First, for the error in your reasoning:
Note that (with $b$ fixed) solving for $P(a) \geq 0 $ does not impose $ D \leq 0$.
That is true only if $P(a) \geq 0 $ for all values of $a$ (which the problem doesn't require).
In this case, substantiated by your observation that $ D = 1+b^4 > 0 $, we know that for each $b$, there will be some positive and negative values of $P(a)$.
To find which are the valid $a$ values, we have to find the roots of the quadratic equation. This is essentially shown by Prasiortle.
Now, for a quick solution:
Write the inequality as $ \frac{2}{b} \geq \frac{ a^2 - 1}{a}$.
Observe that for $ a \geq 2$, we always have $ a < a^2 - 1$.
Hence, for this inequality to hold true, we must have $ \frac{2}{b} \geq \frac{ a^2 -1 }{ a} > 1 $, or that $ b =1$.
Thus, this gives us $ 2a \geq (a^2 -1 )$, or that $ a = 2$.
Now, for my original solution using a different method.
Let $ a = b + k$ where $ k$ is a positive integer.
Then the inequality becomes
$$ 2b + 2k \geq 2b^2k + bk^2. $$
From here, since
It follows that we must have $k = 1$. This then gives us
$$2b + 2 \geq 2b^2 + b $$
and solving the quadratic shows us that the only positive integer solution is $ b = 1$.
Hence, the only solution is $ (a, b) = (2, 1)$.